Add Memory

Store new memories by sending conversation messages. Audaxic automatically extracts meaningful facts and stores them as searchable vector embeddings.

Endpoint

POST https://api.memory.audaxic.com/add

Request Body

{
  "messages": [
    "I went to the vet on May 7, 2023",
    "My favorite programming language is Rust"
  ],
  "timestamp": "2023-05-07"
}

Field	Type	Required	Description
messages	string[]	Yes	Array of messages to extract memories from
timestamp	string | null	No	Session date for resolving relative time references

Timestamp modes

Value	Behavior
Omitted / null	Defaults to today's date
ISO 8601 date (e.g. "2023-05-07")	Used as anchor for relative references like "last week"
"no_date"	Stores memories without any resolved date

When you provide a timestamp, the AI engine uses it to resolve relative time references in your messages. For example, if timestamp is "2023-05-14" and the message says "last week", the memory is dated to the week of 2023-05-07.

If no explicit date is mentioned in the message and no timestamp is provided, the memory is stored without a date.

Response

{
  "results": [
    {
      "id": "mem_abc123",
      "memory": "User went to the vet on May 7, 2023",
      "metadata": { "session_timestamp": "2023-05-07" }
    },
    {
      "id": "mem_def456",
      "memory": "User's favorite programming language is Rust",
      "metadata": { "session_timestamp": "2023-05-07" }
    }
  ]
}

Field	Type	Description
results	MemoryItem[]	List of extracted and stored memories
results[].id	string	Unique memory identifier
results[].memory	string	The extracted memory text
results[].metadata	object	Associated metadata including session_timestamp

Examples

With timestamp

cURL:

curl -X POST https://api.memory.audaxic.com/add \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ak_your_api_key" \
  -d '{
    "messages": ["I went to the vet on May 7, 2023"],
    "timestamp": "2023-05-07"
  }'

JavaScript:

const response = await fetch("https://api.memory.audaxic.com/add", {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    Authorization: "Bearer ak_your_api_key",
  },
  body: JSON.stringify({
    messages: ["I went to the vet on May 7, 2023"],
    timestamp: "2023-05-07",
  }),
});

const data = await response.json();
console.log(data.results);

Python:

import requests

response = requests.post(
    "https://api.memory.audaxic.com/add",
    headers={"Authorization": "Bearer ak_your_api_key"},
    json={
        "messages": ["I went to the vet on May 7, 2023"],
        "timestamp": "2023-05-07",
    },
)

data = response.json()
print(data["results"])

Without timestamp (defaults to today)

cURL:

curl -X POST https://api.memory.audaxic.com/add \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ak_your_api_key" \
  -d '{
    "messages": ["I prefer dark mode and use VS Code"]
  }'

Without date resolution

curl -X POST https://api.memory.audaxic.com/add \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ak_your_api_key" \
  -d '{
    "messages": ["User enjoys painting landscapes"],
    "timestamp": "no_date"
  }'

How Extraction Works

When you send messages, the AI engine:

1. Analyzes the content for factual information
2. Extracts distinct facts (each becomes a separate memory)
3. Generates vector embeddings for semantic search
4. Stores memories in your isolated user space

For example, sending "I'm a software engineer who loves cats and lives in New York" might produce three separate memories:

• "User is a software engineer"
• "User loves cats"
• "User lives in New York"